It's actually a probability of 0.48 to hit it in 5 or less tries:
>>> 1-(1-8000.0/65536)**5
0.478446726626669
ie. you're almost as likely to hit lessthan 8000 in 5 tries or less than you are to fail to hit it.
Here's a strategy for winning almost 500 BTC:
win on turn 1 : bet = 71.08115510, spent = 71.08165510, win = 568.67660461, lose = 0.35490577, back = 0.00000000, profit = 497.59494951 (87.793% to lose all)
win on turn 2 : bet = 81.18455783, spent = 152.26671293, win = 649.50778702, lose = 0.40542278, back = 0.35490577, profit = 497.59597986 (77.076% to lose all)
win on turn 3 : bet = 92.72404791, spent = 244.99126084, win = 741.82823108, lose = 0.46312023, back = 0.76032855, profit = 497.59729879 (67.667% to lose all)
win on turn 4 : bet = 105.90374931, spent = 350.89551015, win = 847.27100871, lose = 0.52901874, back = 1.22344878, profit = 497.59894734 (59.407% to lose all)
win on turn 5 : bet = 120.95679999, spent = 471.85281014, win = 967.70131497, lose = 0.60428399, back = 1.75246752, profit = 497.60097235 (52.155% to lose all)
You'll see that there's a 52% chance of losing all 5 bets (a total of 472 BTC), but if any of the 5 bets win, you'll win 497 BTC. That seems like a good deal to me actually - because it feels like 497.6 * 47.845% is bigger than 471.85 * 52.155% (but, of course, it isn't - there's no "good deal" when playing SatoshiDice...)
Also, if you won on the 5th roll, you'd be the 2nd biggest winner in the history of SatoshiDice.
But every roll is a different chance to win/lose no? So technically on the 5th roll you still have a 88% chance to lose.